API Docs

Production::
https://kcmj21es3e.execute-api.us-east-1.amazonaws.com/prod/v1/docs
https://kcmj21es3e.execute-api.us-east-1.amazonaws.com/prod/v1/redoc

This Page Source Code:
https://github.com/Radio-Memory/radiomemory-ai-api-demo

import config as cfg
import requests
import cv2
from PIL import Image
import io
import base64
import matplotlib.pyplot as plt
import pprint

pp = pprint.PrettyPrinter(indent=4)

BASE_URL = "https://kcmj21es3e.execute-api.us-east-1.amazonaws.com/prod/v1"

def encode_image(image):

    buffered = io.BytesIO()
    image.save(buffered, format="JPEG")
    image64 = base64.b64encode(buffered.getvalue())
    return image64.decode()

Authentication

response = requests.post(
        BASE_URL + "/auth/token",
        data={
            "username": cfg.USERNAME,
            "password": cfg.PASSWORD,
        },
    )
token = response.json()["access_token"]
headers = {"Authorization": "Bearer %s" % token}

Panoramics

# The image is converted to grayscale in the backend, so you can convert it before sending in order to save bandwidth.
panoramic_image = Image.open("images/panoramic0.jpg").convert("L")

Dentition Endpoint

Classify the panoramic radiography in either toothless, superior toothless, inferior toothless, mixed or "normal".

dentition_response = requests.post(
        BASE_URL + "/panoramics/dentition",
        json={
            "base64_image": encode_image(panoramic_image),
        },
        headers=headers,
    )

pdata = dentition_response.json()
pdata["entities"] = pdata["entities"][:4]
pp.pprint(pdata)

{   'entities': [   {   'bbox': None,
                        'class_name': 'normal',
                        'contour': None,
                        'line': None,
                        'point': None,
                        'score': 1.0,
                        'tooth': None,
                        'type': 'dentition'},
                    {   'bbox': None,
                        'class_name': 'superior_toothless',
                        'contour': None,
                        'line': None,
                        'point': None,
                        'score': 7.882013930426979e-10,
                        'tooth': None,
                        'type': 'dentition'},
                    {   'bbox': None,
                        'class_name': 'mixed',
                        'contour': None,
                        'line': None,
                        'point': None,
                        'score': 2.5856181778971532e-11,
                        'tooth': None,
                        'type': 'dentition'},
                    {   'bbox': None,
                        'class_name': 'inferior_toothless',
                        'contour': None,
                        'line': None,
                        'point': None,
                        'score': 1.471757177934352e-12,
                        'tooth': None,
                        'type': 'dentition'}],
    'height': 1292,
    'image_hash': None,
    'width': 2444}

fig, ax = plt.subplots(figsize=(10, 10))
ax.imshow(panoramic_image, cmap="gray")
ax.axis("off");

Longaxis Endpoint

Returns a point coordinates for the crown and the root of every tooth in the panoramic image.

longaxis_response = requests.post(
        BASE_URL + "/panoramics/longaxis",
        json={
            "base64_image": encode_image(panoramic_image),
        },
        headers=headers,
    )

Output Sample

pdata = longaxis_response.json()
pdata["entities"] = pdata["entities"][:3]
pp.pprint(pdata)

{   'entities': [   {   'bbox': None,
                        'class_name': '18_0',
                        'contour': None,
                        'line': None,
                        'point': [632.8214285714287, 553.7142857142857],
                        'score': 0.5159812122583389,
                        'tooth': None,
                        'type': 'longaxis'},
                    {   'bbox': None,
                        'class_name': '18_1',
                        'contour': None,
                        'line': None,
                        'point': [611.0, 351.83928571428567],
                        'score': 0.5159812122583389,
                        'tooth': None,
                        'type': 'longaxis'},
                    {   'bbox': None,
                        'class_name': '17_0',
                        'contour': None,
                        'line': None,
                        'point': [752.8392857142858, 594.0892857142857],
                        'score': 0.594242125749588,
                        'tooth': None,
                        'type': 'longaxis'}],
    'height': 1292,
    'image_hash': None,
    'width': 2444}

Visualization

from vis import draw_longaxis_output

dimage = draw_longaxis_output(panoramic_image, longaxis_response.json()["entities"], draw_axis=True)
fig, ax = plt.subplots(figsize=(10, 10))
ax.imshow(dimage)
ax.axis("off");

Panorogram Endpoint

Returns the panorogram curves of the panoramic image.

panorogram_response = requests.post(
        BASE_URL + "/panoramics/panorogram",
        json={
            "base64_image": encode_image(panoramic_image),
        },
        headers=headers,
    )

Output Sample

pdata = panorogram_response.json()
pdata["entities"] = [{**p, "contour": p["contour"][:3]} for p in pdata["entities"][:2]]
pp.pprint(pdata)

{   'entities': [   {   'bbox': None,
                        'class_name': 'ContMand',
                        'contour': [[150, 179], [149, 180], [148, 180]],
                        'line': None,
                        'point': None,
                        'score': None,
                        'tooth': None,
                        'type': 'panorogram'},
                    {   'bbox': None,
                        'class_name': 'CanManDir',
                        'contour': [[386, 547], [385, 548], [384, 548]],
                        'line': None,
                        'point': None,
                        'score': None,
                        'tooth': None,
                        'type': 'panorogram'}],
    'height': 1292,
    'image_hash': None,
    'width': 2444}

Visualization

from vis import draw_panorogram
dimage = draw_panorogram(panoramic_image, [[p["class_name"], p["contour"]] for p in panorogram_response.json()["entities"]])
fig, ax = plt.subplots(figsize=(10, 10))
ax.imshow(dimage)
ax.axis("off");

Metals Endpoint

Returns metal structures found on the panoramic image.

metals_response = requests.post(
        BASE_URL + "/panoramics/metals",
        json={
            "base64_image": encode_image(panoramic_image),
        },
        headers=headers,
    )

Sample Output

pp.pprint(metals_response.json())

{   'entities': [   {   'bbox': [   860.0716552734375,
                                    624.821044921875,
                                    1569.6455078125,
                                    783.9153442382812],
                        'class_name': 'ortodontia',
                        'contour': None,
                        'line': None,
                        'point': None,
                        'score': 0.9565339088439941,
                        'tooth': None,
                        'type': 'metals'},
                    {   'bbox': [   800.822021484375,
                                    469.63043212890625,
                                    1654.4962158203125,
                                    637.3797607421875],
                        'class_name': 'ortodontia',
                        'contour': None,
                        'line': None,
                        'point': None,
                        'score': 0.831466555595398,
                        'tooth': None,
                        'type': 'metals'}],
    'height': 1292,
    'image_hash': None,
    'width': 2444}

Visualization

from vis import draw_bboxes

dimage = draw_bboxes(panoramic_image, metals_response.json()["entities"])
fig, ax = plt.subplots(figsize=(10, 10))
ax.imshow(dimage)
ax.axis("off");

Teeth Segmentation Endpoint

Return a segmentation mask in the form of a contour for each found tooth on the panoramic image.

teeth_segmentation_response = requests.post(
        BASE_URL + "/panoramics/teeth-segmentation",
        json={
            "base64_image": encode_image(panoramic_image),
        },
        headers=headers,
    )

Output Sample

teeth_segmentation_response

<Response [200]>

pdata = teeth_segmentation_response.json()
pdata["entities"] = [{**p, "contour": p["contour"][:3]} for p in pdata["entities"][:3]]
pp.pprint(pdata)

{   'entities': [   {   'bbox': None,
                        'class_name': '18',
                        'contour': [[608, 355], [607, 356], [606, 356]],
                        'line': None,
                        'point': None,
                        'score': None,
                        'tooth': None,
                        'type': 'teeth_segmentation'},
                    {   'bbox': None,
                        'class_name': '17',
                        'contour': [[701, 370], [700, 371], [699, 371]],
                        'line': None,
                        'point': None,
                        'score': None,
                        'tooth': None,
                        'type': 'teeth_segmentation'},
                    {   'bbox': None,
                        'class_name': '16',
                        'contour': [[828, 362], [827, 363], [826, 363]],
                        'line': None,
                        'point': None,
                        'score': None,
                        'tooth': None,
                        'type': 'teeth_segmentation'}],
    'height': 1292,
    'image_hash': None,
    'width': 2444}

Visualization

from vis import contour2mask, draw_masks

contours_list = [[p["contour"]] for p in teeth_segmentation_response.json()["entities"]]
masks = [contour2mask(contour, *panoramic_image.size) for contour in contours_list]
dimage = draw_masks(panoramic_image, masks)
fig, ax = plt.subplots(figsize=(10, 10))
ax.imshow(dimage)
ax.axis("off");

Procedures Endpoint

Returns dental proceadures found on the panoramic image.

procedures_response = requests.post(
        BASE_URL + "/panoramics/procedures",
        json={
            "base64_image": encode_image(panoramic_image),
        },
        headers=headers,
    )

Output Sample

pdata = procedures_response.json()
pdata["entities"] = pdata["entities"][:1]
pp.pprint(pdata)

{   'entities': [   {   'bbox': None,
                        'class_name': 'MatRestCoron',
                        'contour': None,
                        'line': [   [632.8214285714287, 553.7142857142857],
                                    [611.0, 351.83928571428567]],
                        'point': None,
                        'score': 0.6624748706817627,
                        'tooth': '18',
                        'type': 'procedure'}],
    'height': 1292,
    'image_hash': None,
    'width': 2444}

Visualization

from vis import draw_procedures_output

dimage = draw_procedures_output(panoramic_image, [e for e in procedures_response.json()["entities"] if e["score"] > 0.5])
fig, ax = plt.subplots(figsize=(12, 12))
ax.imshow(dimage)
ax.axis("off");

Describe Endpoint

The describe endpoint will return the result of all models at once.
The models can be called individually from the endpoints described in the docs.

describe_response = requests.post(
        BASE_URL + "/panoramics/describe",
        json={
            "base64_image": encode_image(panoramic_image),
        },
        headers=headers,
    )

Output Sample

pdata = describe_response.json()
pdata["entities"] = pdata["entities"][:3]
pp.pprint(pdata)

{   'entities': [   {   'bbox': None,
                        'class_name': 'normal',
                        'contour': None,
                        'line': None,
                        'point': None,
                        'score': 1.0,
                        'tooth': None,
                        'type': 'dentition'},
                    {   'bbox': None,
                        'class_name': 'superior_toothless',
                        'contour': None,
                        'line': None,
                        'point': None,
                        'score': 7.882013930426979e-10,
                        'tooth': None,
                        'type': 'dentition'},
                    {   'bbox': None,
                        'class_name': 'mixed',
                        'contour': None,
                        'line': None,
                        'point': None,
                        'score': 2.5856181778971532e-11,
                        'tooth': None,
                        'type': 'dentition'}],
    'height': 1292,
    'image_hash': None,
    'width': 2444}

Anomalies Heatmap Endpoint

Returns a heatmap of the anomalies found on the panoramic image.

anomalies_response = requests.post(
        BASE_URL + "/panoramics/teeth-anomalies-heatmap",
        json={
            "base64_image": encode_image(panoramic_image),
        },
        headers=headers,
    )

# Found classes:
for idx, data in enumerate(anomalies_response.json()):
    if idx > 7:
        break
    if len(data["anomalies"]) == 0:
        continue
    print(f"Tooth {data['tooth_name']}  Anomalies -" + ", ".join([anno["anomaly_name"] for anno in data["anomalies"]]), end=" ")
    print("\n")

Tooth 18  Anomalies -NodP 

Tooth 17  Anomalies -Calc, NodP, Car 

Tooth 16  Anomalies -Calc, NodP, Car 

Tooth 13  Anomalies -Calc 

Tooth 12  Anomalies -Calc, Car 

Tooth 11  Anomalies -Calc, Car 

Tooth 21  Anomalies -Calc, Car 

Visualization

Visualizing the Calc class.

from utils import process_ret_api
from vis import draw_heatmap

heatmap = process_ret_api(anomalies_response.json(), h=panoramic_image.height, w=panoramic_image.width, anomaly2see="Calc")
dimage = draw_heatmap(panoramic_image, heatmap)
fig, ax = plt.subplots(figsize=(10, 10))
ax.imshow(dimage)
ax.axis("off");

Periapicals

# The image is converted to grayscale in the backend, so you can convert it before sending in order to save bandwidth.
periapical_image = Image.open("images/periapical0.jpg").convert("L")

Classify Endpoint

The classify endpoint will return the periapical type among the 14 possible regions.

classify_response = requests.post(
        BASE_URL + "/periapicals/classify",
        json={
            "base64_image": encode_image(periapical_image),
        },
        headers=headers,
    )

Output Sample

pdata = classify_response.json()
pdata["entities"] = pdata["entities"][:4]
pp.pprint(pdata)

{   'entities': [   {   'bbox': None,
                        'class_name': '42-41-31-32',
                        'contour': None,
                        'line': None,
                        'point': None,
                        'score': 0.9996941089630127,
                        'tooth': None,
                        'type': 'periapical_classification'},
                    {   'bbox': None,
                        'class_name': '33',
                        'contour': None,
                        'line': None,
                        'point': None,
                        'score': 0.0002522027643863112,
                        'tooth': None,
                        'type': 'periapical_classification'},
                    {   'bbox': None,
                        'class_name': '43',
                        'contour': None,
                        'line': None,
                        'point': None,
                        'score': 3.3788826840464026e-05,
                        'tooth': None,
                        'type': 'periapical_classification'},
                    {   'bbox': None,
                        'class_name': '24-25',
                        'contour': None,
                        'line': None,
                        'point': None,
                        'score': 6.261640010052361e-06,
                        'tooth': None,
                        'type': 'periapical_classification'}],
    'height': 1621,
    'image_hash': None,
    'width': 1236}

fig, ax = plt.subplots(figsize=(10, 10))
ax.imshow(periapical_image, cmap="gray")
ax.axis("off");

Longaxis Endpoit

Return the coordinates on the longaxis of each teeth found on the periapical image.

Sample Output

longaxis_response = requests.post(
    BASE_URL + "/periapicals/longaxis",
    json={
        "base64_image": encode_image(periapical_image),
    },
    headers=headers,
)

pdata = longaxis_response.json()
pdata["entities"] = pdata["entities"][:3]
pp.pprint(pdata)

{   'entities': [   {   'bbox': None,
                        'class_name': '43_1',
                        'contour': None,
                        'line': None,
                        'point': [135.1875, 1215.75],
                        'score': 0.05661439150571823,
                        'tooth': None,
                        'type': 'longaxis'},
                    {   'bbox': None,
                        'class_name': '43_0',
                        'contour': None,
                        'line': None,
                        'point': [19.3125, 329.265625],
                        'score': 0.2382279932498932,
                        'tooth': None,
                        'type': 'longaxis'},
                    {   'bbox': None,
                        'class_name': '42_1',
                        'contour': None,
                        'line': None,
                        'point': [299.34375, 1063.78125],
                        'score': 0.45180168747901917,
                        'tooth': None,
                        'type': 'longaxis'}],
    'height': 1621,
    'image_hash': None,
    'width': 1236}

Visualization

from vis import draw_longaxis_output

dimage = draw_longaxis_output(
    periapical_image, longaxis_response.json()["entities"], draw_axis=True, th=0.001
)
fig, ax = plt.subplots(figsize=(10, 10))
ax.imshow(dimage)
ax.axis("off");

Cephalometry

tele_image = Image.open("images/tele0.jpg").convert("L")

cefbot_response = requests.post(
    BASE_URL + "/cephalometry/points",
    json={
        "base64_image": encode_image(tele_image),
    },
    headers=headers,
)

pdata = cefbot_response.json()
pdata["entities"] = pdata["entities"][:3] 
pp.pprint(pdata)

{   'entities': [   {   'bbox': None,
                        'class_name': 'Nal',
                        'contour': None,
                        'line': None,
                        'model_name': 'cefalometry',
                        'point': [1399.0, 445.0],
                        'score': 0.875812828540802,
                        'tooth': None},
                    {   'bbox': None,
                        'class_name': 'Na',
                        'contour': None,
                        'line': None,
                        'model_name': 'cefalometry',
                        'point': [1330.0, 445.0],
                        'score': 0.9012552499771118,
                        'tooth': None},
                    {   'bbox': None,
                        'class_name': 'Or',
                        'contour': None,
                        'line': None,
                        'model_name': 'cefalometry',
                        'point': [1268.0, 765.0],
                        'score': 0.8242092132568359,
                        'tooth': None}],
    'image_hash': None,
    'model_name': 'cefalometry',
    'output_height': 2304,
    'output_width': 2104}

Detected Cepholometric points

Nal , Na , Or , SO , RO , Cl , S , Si , Sp , Cls , Po , Cli , Ba , Od , Bo , Op , Co , Bpc , Ar , Bac , C , Dc , Go , Me , Pog , Gn , E , B , Id , M , Pm , D , A , Ena , Pli , Enp , KR , Te , Cm , Ptm , PVT , Vasa , Vasp , Vaia , Vaip , Xi , Bar , D6s , Ams , 6s , Ppd , 6i , Ami , A4 , C4 , PAR , Ap , 3s , 3i , Aii , Iii , Iis , Sf1 , Ais , Pogl , Bl , Li , Stm , Ls , Al , Sn , Prn , Pn , EILI , AA , ADS , AD2 , R , AD1 , ADI , POR , D8 , M8 , D7 , V , T , Tuber , Pi , PTVR , LN , C2 , PPM , C2m , C2pi , C3ps , C3as , C3a , C3m , C3pi , C4ps , C4as , C4ai , C4m , C4pi

Visualization

from vis import draw_points

dimage = draw_points(tele_image, cefbot_response.json()["entities"])
fig, ax = plt.subplots(figsize=(14, 14))
ax.imshow(dimage)
ax.axis("off");